[Mitgcm-support] Re: detatfreesurf == deltatracer
mitgcm-support at dev.mitgcm.org
mitgcm-support at dev.mitgcm.org
Wed Jul 9 15:47:47 EDT 2003
Hi,
For the NLFS and specially with real freah water input,
its clear that for consistency, deltaT_freesurf needs to
be set to deltaT_tracer.
By the way, deltaT_freesurf is not the time step of
barotropic mode but only the time step of the free
surface equation d.eta/dt = -Div.U.
the equivalent time step of the barotropic mode is
in fact the square root of deltaT_freesurf*deltaT_Mom
For Linear free surface,
my interpretation is that the implicit method
filters/damps the fast motions (here d.eta/dt) what
cannot be (accurately) represented by the time step,
and conserve the slow motions that the time step
allows to represent accurately.
The part of the free surface that matters for the tracer
is d.eta/dt (= w at the surface, but also a contribution to
w at each level that decreasing linearly to 0 at the bottom)
For the momentum, its mainly the surface pressure gradient.
When deltaT_tracer = 72 * deltaT_Mom: the solver filters/damps
variations of eta that are fast relative to deltaT_freesurf;
if deltaT_freesurf=deltaT_Mom,
w_s = d.eta/deltaT_freesurf can still be too large for the tracer eq
(w_s*deltaT_tracer) given the long tracer time step, even if it is
small enough regarding the implicit method "criteria" (enough damping)
because of the small time step deltaT_freesurf.
So I think that the safe way to use multiple time step is
to give the right (long) time step to the solver
(deltaT_freesurf=deltaT_tracer), so that it will damp enough
w_s according to this "long" time step.
And deltaT_Mom is still used for the
surface pressure gradient term, like it is now.
Alternatively, we could have defined a barotropic time step
and use this one every where in the solve_for_P part of the code,
both for the surface pressure term and the divergence of the flow.
it could be set to the square root of deltaT_tracer*deltaT_Mom
with similar effects compared to deltaT_freesurf=deltaT_tracer.
But I don't think its a good method because its not
consistent with anay other parts of the model (Tracer, momentum,
real fresh water forcing ...).
For rigid lid case, I was thinking (?) that deltaT_freesurf
was not used. Is it right ?
Jean-Michel
More information about the MITgcm-support
mailing list